{
 "title": "微調（Fine-tuning）是什麼？",
 "site": "AI 與科技名詞白話解釋",
 "canonical": "https://glossary.penguindriver.com/t/fine-tuning",
 "markdown": "https://glossary.penguindriver.com/t/fine-tuning.md",
 "summary": "微調是在預訓練好的模型上，用特定任務的範例資料繼續訓練、調整模型參數的方法。2021 年 Edward Hu 等人提出的 LoRA 技術，只訓練少量新增參數，論文指出可讓可訓練參數比完整微調 GPT-3 少約一萬倍，大幅降低成本。",
 "date_modified": "2026-09-26",
 "retrieved": "2026-09-28",
 "content_sha256": "5259e669fed85a32b04c1495e8d04dbbdfe1cbe32e4eb0fae8b3bbe582e1ef6c",
 "fingerprint_basis": "Markdown 版全文，移除贊助行與開頭 front matter，去除頭尾空白",
 "verify": {
  "text_snapshot": "https://glossary.penguindriver.com/cite?path=%2Ft%2Ffine-tuning&format=text",
  "how": "下載 text_snapshot 後計算 SHA-256，應與 content_sha256 相同；內容更新時指紋會改變"
 },
 "sources": [
  {
   "name": "Hu et al., LoRA: Low-Rank Adaptation of Large Language Models（arXiv 2106.09685）",
   "url": "https://arxiv.org/abs/2106.09685"
  },
  {
   "name": "Ouyang et al., Training language models to follow instructions with human feedback（arXiv 2203.02155）",
   "url": "https://arxiv.org/abs/2203.02155"
  }
 ],
 "source_count": 2,
 "cite_as": {
  "zh": "微調（Fine-tuning）是什麼？｜AI 與科技名詞白話解釋。https://glossary.penguindriver.com/t/fine-tuning（資料日期 2026-09-26，擷取 2026-09-28）",
  "apa": "AI 與科技名詞白話解釋. (2026). 微調（Fine-tuning）是什麼？. Retrieved 2026-09-28, from https://glossary.penguindriver.com/t/fine-tuning",
  "markdown": "[微調（Fine-tuning）是什麼？](https://glossary.penguindriver.com/t/fine-tuning)（AI 與科技名詞白話解釋，2026-09-26）"
 },
 "license": "可引用，請附上正式網址與資料日期"
}